Thanks very much for your answer. It is very informative. What fraction of total read pairs is supposed to be unpaired in general in WES data?
Dear,
I used the picard MarkDuplicates option on exome-seq data and I got the following output:
LIBRARY UNPAIRED_READS_EXAMINED READ_PAIRS_EXAMINED UNMAPPED_READS UNPAIRED_READ_DUPLICATES READ_PAIR_DUPLICATES READ_PAIR_OPTICAL_DUPLICATES PERCENT_DUPLICATION ESTIMATED_LIBRARY_SIZE
Unknown Library 16796213 7941846 73335 8044138 1283816 7383 0.324719 21938234
Are the numbers for UNPAIRED_READS_EXAMINED (=16,796,213) and READ_PAIRS_EXAMINED (=7,941,846) as expected or they indicate a problem?
Thanks
1 answer
hi, Yes that indicates something unexpected. For a paired-end data successfully aligned, the unpaired reads should/would be a fraction of the total read pairs examined. In your case its reverse. Also note that the % duplication value is showing as 32.4% How was the FastQC report? Here is an e.g. MarkDuplicates output -
UNPAIRED_READS_EXAMINED 53150
READ_PAIRS_EXAMINED 95066758
UNMAPPED_READS 191324
UNPAIRED_READ_DUPLICATES 20376
READ_PAIR_DUPLICATES 4303567
READ_PAIR_OPTICAL_DUPLICATES 1049365
PERCENT_DUPLICATION 0.045363
ESTIMATED_LIBRARY_SIZE 1326608424
This WES sample had ~95M PE and was aligned with bwa mem. The % dup. here is 4%. But I observe anything upto 10-12% duplication for 60-100x depth WES samples.
hi,
Since this stats is what Picard sees in the aligned BAM file, unpaired reads would mean, of all reads that aligned, those where only one of the mate aligned. If you do samtools flagstat your_bam, then the (mapped) unpaired reads is given as the singleton%.
I have seen this to be mostly < 1%.
Log in to answer this question.